earticle

영어

Nondestructive Testing (NDT) is appropriate for ceramic materials but the image may contain cracks, spiracles, and other foreign substances to form various defects on the surface. Such subtle defects are often neglected on manual visual inspection and it also causes inspector subjectivity problem thus we need am automated computational method to solve the problem. In this paper, we propose a fuzzy logic based method to detect such various defects from the surface of ceramic material. In our method, we first apply fuzzy stretching to enhance the brightness contrast from the input image. Then the fuzzy binarization and upper/lower level search algorithm finds the interval range of defects' existence. Finally ART2 learning algorithm determines different types of defects. The novelty of this paper is avoiding segmentation-identification paradigm and apply simpler image processing technique with fuzzy logic and neural learning in defect identification. In experiment, the proposed method successfully detects poor fusing and tungsten defects.